Detailed explanation of Python np.unique() function

np.unique()Detailed function: return the unique value of the arrayUnique values ​​are sorted from small to large by default

some important parameters

return_index : bool, optional. If set to True, returns the index number of the unique value in the array; otherwise does not return.
Note: The returned array is not the same size as the input array, and only contains indices corresponding to unique values

return_inverse : bool, optional. If set to True, returns the index of the unique value by unique value [for each unique value, the index is the same].

Note: The returned array is exactly the same size as the input array

**return_count: bool, optional. If set to True, returns the number of unique values ​​in the array according to the unique value [for each unique value, the index is the same].

Note: The returned array is exactly the same size as the input array

# Case 1
np.unique([1, 1, 2, 2, 3, 3])
# 得到唯一值数组
array([1, 2, 3])
# 对于多维数组而言，如果不指定axis，那么将多维数组看作一维数组进行唯一值寻找
a = np.array([[1, 1], [2, 3]])
np.unique(a)
array([1, 2, 3])
# 多维数组，指定axis，将axis下的看作一个整体进行唯一值search
# 对于二维数组，axis=0既是寻找完全不同的行
# 对于二维数组，axis=1既是寻找完全不同的列
a = np.array([[1, 0, 0], [1, 0, 0], [2, 3, 4], [2, 3, 4], [1, 4, 6]])
np.unique(a, axis=0)
array([[1 0 0], [1 4 6], [2 3 4]])

# Case 2
a = np.array(['a', 'b', 'b', 'c', 'a'])
u, indices = np.unique(a, return_index=True)
u
# 返回唯一值
array(['a', 'b', 'c'], dtype='<U1')
indices
# 返回唯一值对应的index,由于array只有三个唯一值，因此只有三个index
array([0, 1, 3], dtype=int64)
# 利用唯一值index提取唯一值
a[indices]
array(['a', 'b', 'c'], dtype='<U1')

# Case 3
a = np.array([1, 2, 6, 4, 2, 3, 2])
u, indices = np.unique(a, return_inverse=True)
u
# 返回唯一值
array([1, 2, 3, 4, 6])
indices
# 返回唯一值在原数组的index
array([0, 1, 4, 3, 1, 2, 1], dtype=int64)
# 利用【唯一值在原数组的index】数组得到原数组
u[indices]
array([1, 2, 6, 4, 2, 3, 2])
# Case 3
a = np.array([1, 2, 6, 4, 2, 3, 2])
values, counts = np.unique(a, return_counts=True)
values
# 返回唯一值
array([1, 2, 3, 4, 6])
counts
# 返回每个唯一值的数目
array([1, 3, 1, 1, 1], dtype=int64)、
# 进行原数组的还原，原始数组的顺序不发生变化
np.repeat(values, counts)
array([1, 2, 2, 2, 3, 4, 6])

#  Case 4
# 返回tuple类型的数据
date_uniques = np.unique(date, return_index=True, return_inverse=True, return_counts=True)

Learning link:

• numpy.unique